collaborative-perception

CoAdapt is an LLM-based framework for adaptive collaborative perception in IIoT robotic swarms, reducing communication cost by 38% while maintaining detection precision through dynamic fusion control.

CauseCollab proposes a causal unified and modality-agnostic network to address issues in heterogeneous collaborative perception by disentangling semantic factors from modality-specific confounders, achieving state-of-the-art performance on benchmark datasets.